On today’s episode, Kunle is joined by Oliver Edholm, Co-Founder & CEO of Depict, a Stockholm-based start-up that provides AI-based product recommendations that integrate with any eCommerce store.
A ton of products in your eCommerce store will either keep your growth at a stagnant state or even slow down sales. With hundreds in your catalog, it can hinder a customer’s purchase experience. We can’t have that now, can we? Maybe you didn’t know but there is a way to leverage your current SKUs and create flawless product recommendations.
What you need is to turbocharge your product recommendation engine with Depict. With Depict as your fuel and artificial intelligence as your driver, you can leverage your catalog and create stellar product recommendations based on your user data.
In this episode, Kunle and Oliver talk about how artificial intelligence solves eCommerce product recommendation challenges. You will get to hear about what current product recommendation engines are missing. This is a great episode for business and brand owners looking to create loyal customers and increase average order value.
Here is a summary of some of the most important points made:
On today’s interview, Kunle and Oliver discuss:
The eCommerce GrowthAccelerator Mastermind Facebook Group has just launched.
It is a community…
✔️ for founders and experts passionately involved in eCommerce
✔️ for the truly ambitious wanting to make an impact in the markets they serve
✔️ for those willing and open to help and share with other members
Here is where to apply to join the Facebook group
This episode is brought to you by:
The 2X eCommerce Podcast Show is brought to you by Brightpearl, the number one Retail Operating System for e-commerce, multichannel retail brands and wholesalers who want to Grow Fearlessly.
Lightning 50: Does your fast-growing online retail business stand out from the pack?
Have you taken huge growth leaps in the last year?
If so – step up and be in the running for Brightpearl’s Lightning 50. The definitive list of the fastest growing online businesses in the UK and US.
This episode is brought to you by Klaviyo – a growth marketing platform that powers over 25,000 online businesses.
Direct-to-Consumer brands like ColourPop, Huckberry, and Custom Ink rely on Klaviyo.
Klaviyo helps you own customer experience and grow high-value customer relationships right from a shopper’s first impression through to each subsequent purchase, Klaviyo understands every single customer interaction, and empowers brands to create more personalized marketing moments.
Find out more on klaviyo.com/2x.
This episode is brought to you by Gorgias, the leading helpdesk for Shopify, Magento and BigCommerce merchants.
Gorgias combines all your communication channels including email, SMS, social media, livechat, and phone, into one platform.
This saves your team hours per day & makes managing customer orders a breeze. It also integrates seamlessly with your existing tech stack, so you can access customer information and even edit, return, refund or create an order, right from your helpdesk.
Go to Gorgias.com and mention 2x ecommerce podcast for two months free.
This episode is brought to you by Recharge, the leading subscriptions payment solution for Shopify merchants.
Recharge helps ecommerce merchants of all sizes launch and scale subscription offerings. Recharge powers the growth of over 15,000 subscription merchants and their communities—turning one-time transactions into long-term customer relationships.
Turn transactions into relationships and experience seamless subscription commerce with Recharge.
Find out more on rechargepayments.com/2x.
In this episode, we’re going to be talking with a 19-year-old founder of a company called Depict. We’re going to be going through how AI can essentially help you with product recommendations. It’s a great episode you do not want to miss.
Welcome to the 2X eCommerce podcast show. This is a podcast dedicated to digital commerce insights for retail and eCommerce teams. Each week, on this podcast, we interview a commerce expert, a founder of a digitally native consumer brand, or a representative from a best-in-class commerce SaaS product with a tight remit to give you ideas that you could test right away in your brand. You can improve commerce growth metrics such as conversions, average order value, repeat customers, audience size, and ultimately gross merchant value. We are here to help you sell more sustainably to your customer base and the markets.
This episode is a special one. I was in awe from start to finish with this interview because who I interviewed was more a prodigy than anything else. He’s a 19-year-old high school dropout running an eCommerce SaaS platform called Depict. Depict raised $20 million in funding. He has a team size of about 31. This chap is 19. He’s 19. He’s raised $20 million by solving product recommendation issues using artificial intelligence.
We’ll get into the algorithm. I drill him as to the challenges we face as eCommerce marketers or eCommerce people and how he’s using artificial intelligence to solve product recommendation challenges. He also gives challenges. Their constraint in terms of AI is only as good as the volume of information you give it to learn. It’s a phenomenal interview. He’s a charming chap. He worked in Klarna. He was one of the early employees in Klarna. He’s lived in Singapore. He was a machine learning engineer at Klarna. He joined YCombinator at 17.
This is a phenomenal episode. You’ll be impressed. Towards the end, I talk about mindfulness. He talks about how mindfulness is an important pillar in his life. Of-recording, we go into some questions. I questioned him about mindfulness and meditating. What do you do when you get ideas when you’re meditating? Meditation is all about thoughts coming in and leaving you and you’re observing. He answered a question on that one. This is a phenomenal episode. Enjoy this episode. It’s a special one. Do let me know what you think. At that age, he understands the challenges in commerce, which I found phenomenal. It’s interesting. Enjoy this episode. I shall catch you on the other side. Cheers.
Oliver, it’s a pleasure having you on the 2X eCommerce podcast show. Welcome.
There’s so much to say about you. You’re 19 years old. You’re running an eCommerce SaaS platform based on AI. You’re a YC graduate. You have 47 members on your team. You’ve raised over $20 million. It’s incredible. I’ve said a little. Would you mind giving the readers your story to date as a 19-year-old founder of an eCommerce platform based out in Stockholm?
There’s a lot to talk about. I’m one of those people who got into programming quite early. I like computers. When I was younger, I played a lot of Minecraft. If you have kids and they play Minecraft, that’s good. If they get bored of it, they can go in and change the code behind Minecraft. It’s a great segue to learning coding. That’s how I learned coding.
After I did that, I realized that you can do more things to change Minecraft with programming. I made my first smartphone app. You get the bug of building things and doing things that previously seemed like things in the adult world. You realize, “I can do stuff in the adult world. Age doesn’t matter to the same extent.” That’s how I got into coding.
When I was around 13 to 14 years old, I read a book about AI, Artificial Intelligence. I can see that as a big thing for me since that book explained why AI and its development could probably be the most impactful thing for humanity ever. It’s possible that we could create an AI that can do most of the tasks that humans do today. I found that inspiring.
What’s the name of the book?
Superintelligence by Nick Bostrom. It’s a little bit nerdy but it worked for me. By reading that book, I decided that this is my life mission. I want to work on this. There can be things that can go wrong, for sure. It’s most probable that things will go wrong. Either way, you should probably be involved and make the best out of it. I dropped a lot of school-related things and went all-in on that.
How old were you at the time?
About 13 or 14 years old. I am obsessive. I took these online university courses to learn the fundamental algorithms. I used to build stuff. I learned by doing things and building. I build stuff and then I uploaded my AI code to the internet. Through that, when high school was approaching, I got in touch with Klarna. They’re being in the eCommerce world.
Back in 2017, I interviewed the founder of Klarna, Niklas Adalberth. At the time, he was young. I was amazed. You guys both come from Sweden. Is there a thing about the environment as to supporting young tech entrepreneurs that are out in Sweden?
I don’t have anything specific I could pinpoint. We have a good societal system where you’re taking a risk and failing with your company. There’s a system where you can fall back like a social welfare system. That could probably be something around that. Otherwise, I don’t have anything specific. We’re quite IT proficient. We’re early adopters.
You were building, you got the attention of Klarna, and then what happened?
I took this summer internship with them and then they got impressed or something with what I could do. I started working at Klarna with the AI research team there. Instead of going to these lectures about some history stuff or whatever, I was doing stuff in the real world and doing what I wanted to do even after university. I felt that I want to double down on this. I dropped out of high school and went all-in on working on AI. It was during that time that the idea of Depict came about, the company I have today. You had one foot within the latest developments in AI but also one foot in eCommerce and seeing the needs of the merchants.
If you look at Amazon, for instance, they have a great track record of building an organization where they can apply the latest AI research to various parts of the organization and see a quantifiable business impact as a result of that. If you look at product recommendations, which is what we had Depict do, there was a McKinsey report that said that 35% of Amazon sales can be somehow attributed to the recommendation engine. If you look at talks with Jeff Bezos, he’s always bullish about the recommendations and what he always talks about it.
Amazon has got a great impact on that. If you look at the rest of the industry, it’s not like that at all. It’s way behind. That’s primarily due to the fact that modern AI requires a lot of data and in the case of product recommendations, a lot of user data to function well. Most industries don’t have as much data as Amazon. You have a big problem. The idea of Depict is, can we provide a product recommendation engine that requires fewer user data by looking at other data sources? For instance, the product catalog information. Through that, democratize Amazon-level product recommendations to everyone in eCommerce and then eventually do much more than product recommendations.
I started this company at 17. It’s pretty special. It’s one thing building the product and then it’s one thing to sell it, for instance. I didn’t have as many connections and so forth. That was an interesting learning experience doing other things outside. It’s been going pretty well so far. We raised over $20 million. We’re over 37 people at the company. Most of our customers are into Nordics. We have Staples, Office Depot, and Björn Borg the tennis brand. We’ve got more quick commerce companies using us. It’s going pretty well.
The heart of all you do at Depict is artificial intelligence, AI. What’s your take on personalization and its convergence with AI?
Personalization is a huge buzzword in this industry. A lot of times, people use the word personalization without knowing exactly what they mean. You can do great personalization with the use of AI but you don’t have to use AI to do personalization. We’ve done this with some of our customers. What if you configure a rule so that if someone enters your product detail page from Google Shopping, you make the recommendation bar a little bit more visible?
We found that if you don’t find the right product directly after you click on your Google Shopping app, you close the tab and go to a competitor. That’s a simple form of personalization but efficient. It’s simple rules like that where you’re informed by what the customer wants and needs. That can work great as well. That’s how I would say it.
You mentioned Amazon and a few other quite huge organizations at scale. Amazon has billions if not millions of touchpoints. You have first-party data and zero-party data to hone in on the personalization for other players or other commerce operators. What is the threshold from the perspective of catalog size, variety, as well as traffic? This is individual user data and not necessarily just traffic. I could potentially browse on my mobile and later on my desktop. From a user profile standpoint and catalog size standpoint, when do retailers start to benefit from the impact of AI-driven product recommendations?
It’s true that the bigger the catalog you have, the more leverage you get up to, especially product recommendations. You can do some personalization with smaller catalog sizes and still get a great impact after that. If you have 50 products on your site, you can always find what you want in somewhat a relatively easy way. If you have 1,000-plus products, suddenly, there’s so much you can do. We have a minimum threshold of 300 products for someone to work with. The more products, the better. Our largest customer has over 16 million SKUs.
If our recommendation algorithm wants to recommend you something but the product is not in your catalog, then there’s not much you can do. You need a playing room to do stuff. That’s something relevant. You mentioned the traffic. We help convert the customer on the site, make them loyal, and increase the average order value. If you don’t have traffic, then you probably should sell your traffic problem first.
It’s also relative. If you have sizable traffic and you think you can improve the experience for the customer, using something like Depict becomes relevant. If you try to use the traditional methods to do product recommendations, it requires a lot of user data. What that specifically means is people who bought this product also about this product. People who watch this product also watch that product. If you try to do that and you don’t have as much traffic and user data where it has happened before, then the recommendations become pretty weird.
If someone happened to buy this product with this product and maybe you’re a niche drop shipping business where you have a lot of product and less traffic, then the recommendation becomes weird. What Depict can do is we can take into account much more context, especially the product information. For instance, understanding what you are selling here. Through that, even though you don’t have any historical transaction data or user data, we can still do great recommendations.
Looking at your website, there seem to be different touchpoints for recommendations. I’m looking at the front page, for instance. On the products page, the recommendations are there. I want to speak about the category pages. Is it a recommendation of what they see? The most likely products are listed on my product catalog page that I’m likely to click on or is this another block within the category page in which you then say, “These are the products we recommend.”
What you mentioned were the category listings. You can sort by price, relevance, and popularity, and then you can filter for various tags and categories. This isn’t as normal for customers. Some customers asked about having some recommendation carousel on top or below that listing where they can, for instance, highlight various products.
Let’s say they have less of a sophisticated category listing algorithm. Depict has been proving that we provide a lot of value through our AI. We can provide more relevance and personalize what you’ve seen before. Maybe you want to push products that are important for your business, higher-margin products, private label products, newly launched products, and so forth. That specific carousel is less normal for customers.
You mentioned using your experience from product recommendations to other things. Do you want to shed a bit more light on the other insights, AI-driven product recommendations can deliver from a value perspective?
When we recommend products, by default, we need to know something about your user and your products to start with. There are some insights there. We are working on figuring out how we can use those insights in our portal, for instance, to display those things. Maybe you have a poor click-through rate on this product attribute in general. Maybe when you show this product to this customer, they like it.
Probably, if you figure out the target marketing campaign for that customer with the specific product, you can get a lot of leverage out of that. Those things are interesting. In terms of other insights, there are a lot of things we can do. Our customers have been asking for more things like that. That’s something we’re working on incorporating.
With Depict, another question I wanted to ask was about the AI components. You’ve made a point that you’re not user-driven in terms of making the product recommendations. You are taking in what happens on the site, traffic patterns. Are there any other data points your AI system picks and stores to make the perfect product recommendation for each session?
The number one existing product recommendation engines have been missing is incorporating the product information in a good way. Understand what products are you selling. Most product recommendations see products as IDs or SKU IDs. They say, “This ID correlates a lot with this ID. For some reason, we should recommend it.” We have tons of more context since we apply AI to the images. What’s the pattern of this dress? What brand does it come from and so forth? We combine that with the little user data they have. We’re still user-driven. You have so much more context you can recommend.
Your AI can identify that this is an iPhone. All of a sudden, your AI knows what the entity is, the iPhone, and its potential uses. When you buy an iPhone, the next thing you want to buy is a case or an extra charger. You have that context of what exactly it is rather than an ID and then you combine that with all the other data points. That’s clever.
You maybe have your tags and so forth. We go far beyond what’s in your product feed. We go in and analyze, on a much more subtle level, what’s in the image and what’s in the text you write? We get the more subtle nuances, which aren’t explicitly there.
You’re training an AI to be almost like a shopping assistant. It’s saying, “You got this blazer from our store. You’re looking at it. You’re trying it on. These pair of shoes will work.” It’s something like that. You now have the full context of what exactly that blazer is and the color. It’s learning from private industry as well as its utility.
That makes a ton of sense. Another question I had was about integration. A lot of SaaS companies that come on the show tend to say, “At the moment, we’ve launched for Shopify Plus.” They’re built for Shopify initially. They then start to extend. On your website, you have Shopify, Storm, PrestaShop, Salesforce, Episerver, BigCommerce, Magento, WooCommerce, and many more. You’ve been around for over one year. How have you been able to cater to such a spread from an eCommerce platform standpoint?
What we realize pretty early when talking to these eCommerce merchants was that one of the biggest problems they have is IT capacity of some sort. You know what you want to do but you have a hodgepodge of systems and you need someone who knows how to code to integrate this. For us, that’s frustrating when we know that we provide a fantastic service. If you can get it live on the site, we’ll know it works.
We’ve invested a ton of products. We built our organization around ensuring that the integration is as seamless as possible. We are completely platform-agnostic. If you build your own in-house thing, we can integrate with you as well. The only thing you need to integrate is add a script or a tag to your website, give us a link to a product feed, share your account with us through Google Analytics, and then you’re done. We ingest a recommendation bar to the site. It doesn’t affect the loading speed or anything like that.
How can that happen? We’ve done some pretty innovative things to be able to integrate through this tag. That’s where the secret sauce is. It’s doing most of the bulk of the integration work through the script that you add to the site in the same way you add a Facebook or Google pixel to the site. That means we do a little bit of work from our end to set this up. From your perspective, it’s seamless.
I consider you an eCommerce expert given what you guys are doing. From the number of retailers you’re supporting, what do you think is the number one distinguishing factor that separates a best-of-breed eCommerce operator or eCommerce brand from others?
I want to say it’s the IT capacity of some sort but that’s not what I want to say. What I want to say is, how fast can you make changes to your site and iterate? How fast can you do changes, track the results of that, and have a fast feedback loop around that? That’s something that a lot of eCommerce sites are struggling with.
If you have those foundations in place and can be a data-driven merchant, you can go from, “I have a hypothesis. I can implement the changes to the site or my marketing campaign revenue to prove that hypothesis and measure the results.” If you can go through that cycle efficiently, that’s something that will set you apart among many things. It’s a multifaceted thing, eCommerce. There are many things you can focus on. Shipping and logistics are something extremely important that I touched less on.
On the one hand, that operational agility is driven by data and then taking action, insights, hypotheses, feedback, and doing it over and over again. On the other hand, it’s caring about the customers whether it’s shipping times or saying you do what you do on time. It’s super insightful. Thank you. Before I let you go, you mentioned the book Superintelligence. Is there a book you’re currently reading that you’d like to share? Besides that book, any books that could make a pivotal mindset change for the readers?
I can give you something outside eCommerce and business. I would say a book called Waking Up by Sam Harris. It talks about meditation in quite a secular context. The author has been a monk for two years or something. He has a PhD in neuroscience. He’s a rational and scientific guy. He bridges the worlds of Buddhism, spirituality, meditation, and science in a good way. That book has made me pick up a meditation habit. That has a big impact on me and it can have a lot of impact on others as well.
What impact has meditation had on you?
The biggest thing is you can have much more of an objective/third-party perspective on whatever you’re experiencing, primarily your own thoughts. You can be much less attached to the thoughts. You don’t take yourself as seriously if you get concentrated. What I found was that when a thought comes up into your mindstream, there’s already been a lot of processing in your brain, subconsciously. To the same extent, you are thinking this thought.
It’s almost like a video game. You’re a character in your own video game.
It can be much more of a perspective like that. In entrepreneurship, for instance, you can start to move a little bit faster, you can start to take that risk, and acknowledge that, at the end of the day, things are fine.
How often do you meditate?
I meditate 1.5 hours a day. I’ve been building up to that. I do retreats as well. I was in South Africa before and it’s 10 hours a day for 10 days in a row. That’s an interesting experience.
What time of the day do you set aside for meditation at the moment?
In the morning, it’s roughly half an hour after I wake up. I do a lot of meditation to build up my concentration so that I can stay on an object of focus for a longer period of time. That’s what I’m focusing on. That’s helped me also be more focused at work.
Oliver, we could go on and on. By far, you’re one of the most interesting guests I’ve had on the show thus far. I wish you the best. You already have some more traction anyway from a client perspective, employee perspective, headcounts perspective, and also your raise. Thank you so much for coming on the 2X eCommerce podcast show. For those people who want to find out more about Depict, it’s Depict.ai. Are you active on any social media platforms?
You can find me on LinkedIn.
We’re already connected on LinkedIn, you’re Oliver Edholm. You could search for Oliver on LinkedIn. Thank you so much. I appreciate this conversation. Cheers.